BidirLM/laion_audio_contrastive

BidirLM x Laion 音频数据集是一个多模态对比学习数据集，专注于音频和文本的配对数据。它是BidirLM-Omni集合的一部分，作为用于训练BidirLM-Omni模型的语料库的特定模态子样本，规模在10万到100万样本之间。该数据集旨在支持音频与文本之间的跨模态学习任务。

BidirLM x Laion Audio is a multimodal contrastive learning dataset focusing on paired audio and text data. It is part of the BidirLM-Omni Collection, serving as a specific modality sub-sample of the corpus used to train the BidirLM-Omni models, with a size between 100K and 1M samples. This dataset is designed to support cross-modal learning tasks between audio and text.